github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
Guest37021 timotimo: did rr solve yesterdays heap profile problem? 11:01
timotimo Guest37021: it will, i didn't finish the work yet and then while i was trying to do something "for fun" stumbled upon another bug i recently put into rakudo 11:55
Guest37021 so you got a bonus bug then :) 12:35
timotimo yeah 12:36
thankfully that bug i mentioned hasn't gotten into the recent release
but also, there isn't a test that catches it, which is not so good
jnthn survived his workshop and first talk :)
timotimo yay
jnthn Though still pretty tired and ear-achey 12:37
timotimo i'm still watching your talk, since i got up a little too late; using youtube's DVR-like feature
jnthn Will be next week before I really get to look much at things here.
Guest37021 how does this DVR feature work? 12:43
timotimo *shruuuuug* 12:45
they just immediately store the streamed video on top of livestreaming it to people?
Guest37021 I figured it out 12:46
it was super difficult :-)
timotimo oh
that's what you mean
Guest37021 I was wathching the talk but then someone wanted to get ice cream 12:47
I also managed to find a fromspace for Nine 12:48
*fromspace error
perhaps it's the last one 12:49
timotimo perhaps using nine's patches to poison the fromspace and turning on the different profiling options we have will find another thing or two 13:27
Guest37021 perhaps I should try his stuff out 13:32
timotimo: so how will you stop spesh during GC if that's your plan? 13:34
timotimo it's supposed to already be stopped during GC 13:41
actually, it's the other way around
GC is stopped until spesh joins in, and spesh shouldn't be able to resume work until GC is done 13:42
Guest37021 and what says that a GC is done?
timotimo orchestrate.c has that
a combination of atomic variables and condition variables 13:43
Guest37021 is it possible that github.com/MoarVM/MoarVM/blob/mast...ate.c#L264 is placed too far down in finish_gc() ? 13:48
discord6 <timotimo> Oh, hmm 13:53
discord6 <timotimo> Can you try moving that up a little? 13:56
Guest37021 do you have any specific line in mind? 13:57
around line 194 perhaps? 13:59
discord6 <timotimo> I wonder if before the fixed size safe point is okay 14:00
Guest37021 that would be around line 184 then 14:03
timotimo is anatofuz here? 14:07
anagura, that you? 14:08
anagura timotimo Yes, I'm anagura (a.k.a. AnaTofuZ ) 14:09
(Real name is Takahiro Shimizu)
timotimo hi
I'm on my phone to now which makes typing a little more difficult but I can still answers your questions 14:10
just with some typos from autocorrect
anagura Thaks! 14:13
Jonathan talked about dynamic optimization with MoarVM. I want to know that he was logging (probably) the calling method.
(I'm sorry because I'm not good at English) 14:14
I read interp.c, but I think that this features not implementation in this c source code..? 14:17
timotimo interp.c puts data into logs at different points
I believe some other places that do that are everywhere in the code 14:18
github.com/MoarVM/MoarVM/blob/mast...pesh/log.h the functions at the bottom here are the ones you should search for on the code 14:19
anagura Thank you!! I try read this implementation using lldb! 14:20
timotimo good luck! I did not use lldb yet, I do all my c debugging with gdb 14:23
sometimes ddd
anagura ddd?
I often use lldb, Because I use mac os...! ( when use linux, I use gdb) 14:24
moritz ddd = data display debugger, visual frontend for gdb that shows some more data without you having to poke everything 14:25
anagura I did not know that. Thank you 14:26
timotimo the spesh log functions are designed to be as fast as possible do all of the real work happens in the spesh worker
the threads running bytecode want to have add little overhead from logging add possible
worker.c is the spesh thread 14:27
it waits for a complete log to arrive and analyzes it and optimizes the byte code a everything
anagura MoarVM is rewriting bytecode, isn't it? 14:31
I'm so excited ... !!! (wow funny) 14:32
timotimo it adds new versions to the old ones
the original is kept
anagura Ohhhhh!!! I see!!
timotimo hey Guest37021 can you run a small program and gist the spesh log? 14:33
Guest37021 sure 14:35
timotimo anagura: the spesh log gives a lot of insight into how spesh operates 14:36
anagura OK , I will try look spesh log. 14:38
timotimo it should arrive shortly
discord6 <timotimo> I should mention that what I call spesh log there is what SRC/spesh/dump.c creates 14:42
<timotimo> But you activate it with the environment variable MVM_SPESH_LOG 14:43
<timotimo> Btw anagura you can also chat here with discord if you have that already 14:44
anagura Now, I try install latest Perl6 ( execute z init ) 14:53
timotimo oh z script?
cool
Guest37021 the smallest spesh log you can get seems to be over 1 meg in size 14:55
timotimo ha, oops
maybe bad for gist's web interfacw
but can you post the "raw" url?
Guest37021 I tried perl6 -e ''
timotimo that can be curl'd or even opened in the browser 14:56
yeah the spesh log includes lots of stuff from the compiler
Guest37021 are you even allowed to post files of that size
timotimo probably
Guest37021 3200 lines :) 14:57
Guest37021 32k lines I mean 14:57
discord6 <timotimo> It surely compresses well 14:58
Guest37021 can I post a binary as a gist 14:59
discord6 <timotimo> Using got, probwbly 15:00
<timotimo> Got clone, got add, got commit, got pish 15:06
<timotimo> Autocorrect, go!
Guest37021 timotimo: gist.githubusercontent.com/dogbert...tfile1.txt 15:12
timotimo oh no, anagura left :( 15:13
haha 15:16
that spesh log only contains a single frame being specialized
and another had its "before" state dumped when the process shut down 15:17
brrt \o 16:51
timotimo o/ 16:52
Geth MoarVM: d189b29bc3 | (Ben Davies)++ | 11 files
Make socket family handling portable

MVM_io_resolve_host_name assumes a 64-bit system where the values of the AF constants are the same as those on Linux. This breaks domain name resolution on OSes where the values of AF constants aren't the same as Linux's, such as FreeBSD. In addition, Rakudo never set the family to the upper bits of the port passed to the function, meaning that its ... (8 more lines)		 18:06
MoarVM: 352ae27e4c | (Aleks-Daniel Jakimenko-Aleksejev)++ (committed using GitHub Web editor) | 11 files
Merge pull request #1126 from Kaiepi/portable-socket-families

  [IP6NS Grant] Make socket family handling portable
dogbert17 timotimo: moving the calls to MVM_profile_dump_instrumented_data(tc) and MVM_profile_heap_take_snapshot(tc) from lines 263-4 to github.com/MoarVM/MoarVM/blob/mast...ate.c#L184 seems to help 18:21
Kaiepi i'm gonna try to refactor the entirety of src/io/asyncsocketudp.c out of existence for my networking grant at some point so async unix sockets don't triplicate code once i implement them 18:28
that's not gonna be happening for a while, though. at least two months 18:29
ugexe why not do it outside of your grant? 19:08
by adding scope to your grant you make it harder for yourself to deliver everything you claimed
if you don't add the scope to your grant (and just do it) your grant is still the same size, but you don't lose out if you don't hit your targets
brrt \o 21:32
pamplemousse o/ 21:33
brrt pamplemousse: are you still stuck? 21:34
timotimo .o( thunderstuck! ) 21:38
pamplemousse I am substantially less stuck than before. There was an issue with how I was saving the module files, which was causing some trouble, but I've fixed that. I've been following up on what ugexe said about approaching the modules issues: colabti.org/irclogger/irclogger_lo...-08-06#l51 21:39
But as far as I can tell, I think I am actually zipping up all the necessary dependencies now, which is good. 21:40
brrt yay
:-)
Kaiepi ugexe, that may be a better idea. i'm already making a lot of changes for the time i gave myself as it is 22:03

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!